]> dreyeck.freedombox.rocks Git - topic-maps.git/blob - docs/slice-16a-validation.md
Integrate qualified XML anyType values into the XTM reader
[topic-maps.git] / docs / slice-16a-validation.md
1 # Slice 16A — General Item Identifiers: validation
2
3 Base: `8c13acc6e96ced1b14d8ec733f701ac788911720`; local branch
4 `codex/ctm-slice-16a`. The source hashes in `reports/slice-16a/execution.json`
5 identify the actual implementation, tests and documentation used by the run.
6
7 ## Reproduce
8
9 From the repository root, using the existing development SBCL:
10
11 ```sh
12 python3 scripts/validate-slice-16a.py --sbcl /nix/store/hlsiblxpxr1hax3dw9x8znd6zy097ddj-sbcl-2.5.4/bin/sbcl
13 ```
14
15 The validator executes `scripts/test.sh` in a fresh SBCL 2.5.4 process, with no
16 user initialization or user ASDF registry, under an OS sandbox denying network
17 access. It retains every existing suite and independent report checker, adds
18 the Item Identifier suite and independent XML checker, then executes the
19 documented inspection example in a second isolated process. No version switching,
20 checkout relocation, FASL-cache changes or HyperDoc loading are part of this slice.
21
22 `execution.json` records exact commands, UTC execution interval, individual
23 suite totals, total Lisp checks, source/artifact SHA-256 values and exit status.
24 `tests.log` retains the complete output. `identifiers.json` contains individual
25 check labels and seven manual model/baseline observations. `manual-model-actual/`
26 contains their actual CXTM bytes. `describe.log` is the inspection output.
27
28 The recorded SBCL 2.5.4 run passes **9,728 Lisp checks**, comprising the retained
29 9,484 checks and 244 new Item Identifier checks, with zero assertion failures.
30 The Python unit tests, 16,294 NFC oracle comparisons and all independent report
31 checkers pass. The two known imported DateTime baseline FAILs remain separate.
32
33 ## Acceptance evidence
34
35 - Empty and nonempty sets for all seven kinds; idempotence; copied input/output
36 strings and list storage.
37 - Every directed cross-kind collision, distinct non-Topic items of the same
38 kind, legacy constructor routes, and unchanged model snapshots on rejection.
39 - Topic identity merging; Name/Variant/Occurrence/Association/Role identifier
40 unions; old handles used again after replacement; stored Parent/Type/Player
41 references and Scope invariants.
42 - Role-type and Theme merges, Association merging with Role metadata, existing
43 equal Names, and atomic rejection of identifier-induced Variant Scope collapse.
44 - Identifier insertion order and object construction order independence; pure
45 serialization, including retained forward links; no use of CXTM NFC output
46 for TMDM identifier equality.
47 - Seven hand-built TMDM models match the unchanged original XTM2 CXTM baselines.
48 The independent Python checker also verifies all seven XML positions,
49 identifier ordering and Topic references. This does not claim an XTM reader.
50
51 The validator checks every retained suite total against Slice 15, compares all
52 83 successful original CTM/CXTM outputs and the known discrepancy outputs, and
53 compares retained semantic observations. The complete 273-contract statuses
54 remain **93 PASS, 2 FAIL, 0 UNSUPPORTED, 178 NOT-RUN**. The FAILs are the existing
55 `occurrence-datetime2.ctm` and `occurrence-datetime-timezone2.ctm` normative
56 baseline discrepancies. No newly implemented CTM syntax or imported contracts
57 are claimed. All pinned corpus/reference bytes are protected by checksum checks.
58
59 ## Boundaries
60
61 No Reifier/Reified properties, CTM `~`, Reification-induced merging, XTM reader,
62 Templates, Include/Mergemap, cross-document merging or HyperDoc integration.
63 The historical Topic-SI/Occurrence-II guard remains an explicitly tested,
64 documented restriction until Slice 16B; no lexical coincidence is presented as
65 automatic Reification. See `reading-item-identifiers.md` for API and clauses.